\begin{table}[h!]
    \centering
    \caption{Bootstrap Estimates of Citywide Supply Response and Marginal Fiscal Cost, \\ Alternative Specifications}
    \label{tab:bootstrap_ests_lpm_probit}
    \begin{tabular*}{\textwidth}{@{\extracolsep{\fill}}lccc@{}}
\toprule  \addlinespace
            & RE Logit           & Probit                & LPM               \\    \addlinespace \cmidrule{2-4} 
                     & (1)                        & (2)                         \\   \midrule \addlinespace
Supply Response     & 0.57***   & 0.53***                        & 0.59***                           \\
                  & (0.06)             & (0.05)                      & (0.06)                           \\  \addlinespace
Marginal Fiscal Cost & \$1,751,636***  & \$2,480,480***                        & \$1,627,548***                              \\
               & (318,971)                  & (419,883)                      & (339,004)                      \\  \addlinespace \bottomrule
\end{tabular*}
	\begin{spacing}{1}
\begin{tablenotes}
      \item \footnotesize \textit{Notes:} This table presents citywide supply responses and citywide marginal fiscal costs. In the first row, I report estimates of the citywide supply response of onsite inclusionary units to changes in buildings' 421-a incentive. In the second row, I report marginal fiscal costs per inclusionary unit per year. In Column 1, 2, and 3, my estimates are from respectively a random-effects logit specification, and a linear probability model (LPM) specification. The random effects are Neighborhood Tabulation Area intercepts. All specifications assume building features are exogenously determined and include fixed effects for borough and year as well as lot and block controls. Standard errors are computed by a cluster-bootstrap at the level of Neighborhood Tabulation Areas. $\sym{*} = p < 0.10$, $\sym{**} = p < 0.05$, $\sym{***} = p < 0.01$.
      \end{tablenotes}
      \end{spacing}
\end{table}
